# DeepSeek-V3 vs Grok 4.6

> Grok 4.6 is the stronger model overall, scoring 56.9 to 39.5 on the Noometry Index. DeepSeek-V3 costs 7.4× less per token, which makes it the better buy when Grok 4.6's lead doesn't matter for your workload.

- Canonical page: https://noometry.com/compare/deepseek-v3-vs-grok-4-6
- Last updated: 2026-10-11
- Shared benchmarks: 26

## Summary

- They share 26 benchmarks with published results for both. DeepSeek-V3 scores higher in 0 categories and Grok 4.6 in 8 categories; 8 gaps are clear of the uncertainty.
- The widest gap is in reasoning, where Grok 4.6 leads 61.4 to 20.5.
- The biggest single-benchmark swing is OTIS Mock AIME 2024-2025: 37.8% for DeepSeek-V3 and 99.2% for Grok 4.6.
- DeepSeek-V3 is cheaper at $0.24 / $0.90 per million input/output tokens, against $2 / $6 for Grok 4.6.
- Grok 4.6 accepts more context: 500K tokens versus 164K.
- DeepSeek-V3 has downloadable open weights; the other is API-only.

## Snapshot

| | DeepSeek-V3 | Grok 4.6 |
|---|---|---|
| Provider | DeepSeek | xAI |
| Noometry Index | 39.5 | 56.9 |
| Rank | 166 | 21 |
| Context | 164K | 500K |
| Input $/M | $0.24 | $2 |
| Output $/M | $0.90 | $6 |
| Weights | Open | Proprietary |

## Coding

- DeepSeek-V3: 42.3 (#106)
- Grok 4.6: 58.5 (#16)

| Benchmark | DeepSeek-V3 | Grok 4.6 |
|---|---|---|
| SciCode | 35.8% | 56.5% |
| WeirdML | 36.1% | 67.3% |
| LMArena Coding | 1368 | 1465 |
| DeepSWE | — | 67.5% |
| FrontierCode | — | 48% |
| Aider Polyglot | 55.1% | — |
| CursorBench | — | 41.4% |
| LMArena WebDev | — | 1617 |
| FrontierSWE | — | 25.3% |
| BigCodeBench Instruct | 50% | — |
| LiveBench Coding | 70.9% | — |
| BigCodeBench Complete | 62.2% | — |
| ALE-Bench | — | 1,508 |
| HumanEval+ | 86.6% | — |
| MBPP+ | 73% | — |

## Agentic & Tool Use

- DeepSeek-V3: —
- Grok 4.6: 39.4 (#27)

| Benchmark | DeepSeek-V3 | Grok 4.6 |
|---|---|---|
| APEX-Agents | — | 65.3% |
| GDP.pdf | — | 17.2% |
| METR Time Horizons | 49.6% | — |
| Vending-Bench 2 | — | 9,047 |

## Reasoning

- DeepSeek-V3: 20.5 (#236)
- Grok 4.6: 61.4 (#20)

| Benchmark | DeepSeek-V3 | Grok 4.6 |
|---|---|---|
| SimpleBench | 27.2% | 75.9% |
| CritPt | 0% | 19.7% |
| LMArena Hard Prompts | 1365 | 1447 |
| DTBench | 64.8% | 97.3% |
| LMCA | 15.5% | 48.5% |
| Epoch Capabilities Index | 135.94 | 156.44 |
| ARC-AGI-2 | — | 67.1% |
| Kagi LLM Benchmark | 52.3% | — |
| NYT Connections (extended) | — | 80% |
| ARC-AGI-1 | — | 87.5% |
| Chess Puzzles | — | 40% |
| EBR-Bench | — | 30.5% |
| LiveBench Reasoning | 65.8% | — |
| Mystery Game Puzzles | — | 34% |
| LiveBench Data Analysis | 60.9% | — |
| BIG-Bench Hard | 87.5% | — |
| ForecastBench | 59.1 | — |
| HellaSwag | 88.9% | — |
| LiveBench | 66.9% | — |
| PIQA | 84.7% | — |
| WinoGrande | 85.2% | — |

## Math

- DeepSeek-V3: 32.1 (#219)
- Grok 4.6: 67.0 (#24)

| Benchmark | DeepSeek-V3 | Grok 4.6 |
|---|---|---|
| OTIS Mock AIME 2024-2025 | 37.8% | 99.2% |
| LMArena Math | 1373 | 1423 |
| FrontierMath (Tiers 1-3) | — | 66% |
| FrontierMath Tier 4 | — | 31.7% |
| ProofBench | — | 51% |
| Omni-MATH | 40.3% | — |
| LiveBench Math | 73.5% | — |
| MATH Level 5 | 75.5% | — |
| FrontierMath (Feb 2025 set) | 1.7% | — |

## Knowledge

- DeepSeek-V3: 37.5 (#155)
- Grok 4.6: 63.3 (#20)

| Benchmark | DeepSeek-V3 | Grok 4.6 |
|---|---|---|
| GPQA Diamond | 67.6% | 94% |
| LMArena Expert | 1351 | 1467 |
| SimpleQA Verified | — | 49.3% |
| MMLU-Pro | 72.3% | — |
| Confabulations | 26.1% | — |
| Vectara Hallucination Rate | 6.1% | — |
| GPQA (HELM) | 53.8% | — |
| ARC (AI2) Challenge | 95.3% | — |
| MMLU | 87.2% | — |
| TriviaQA | 82.9% | — |

## Multimodal

- DeepSeek-V3: —
- Grok 4.6: 43.6 (#23)

| Benchmark | DeepSeek-V3 | Grok 4.6 |
|---|---|---|
| LMArena Vision | — | 1263 |
| Blueprint-Bench 2 | — | 33.2% |
| Furniture Assembly | — | 40% |
| LMArena Document | — | 1452 |

## Multilingual

- DeepSeek-V3: 48.5 (#143)
- Grok 4.6: 53.0 (#74)

| Benchmark | DeepSeek-V3 | Grok 4.6 |
|---|---|---|
| LMArena Non-English | 1358 | 1420 |
| LMArena Chinese | 1391 | 1480 |
| LMArena French | 1385 | 1461 |
| LMArena German | 1374 | 1431 |
| LMArena Japanese | 1333 | 1376 |
| LMArena Korean | 1319 | 1397 |
| LMArena Russian | 1373 | 1422 |
| LMArena Spanish | 1358 | 1404 |

## Instruction Following

- DeepSeek-V3: 72.8 (#130)
- Grok 4.6: 75.4 (#63)

| Benchmark | DeepSeek-V3 | Grok 4.6 |
|---|---|---|
| LMArena Instruction Following | 1345 | 1431 |
| LiveBench Instruction Following | 81.5% | — |
| IFEval | 83.2% | — |

## Long Context

- DeepSeek-V3: 34.0 (#253)
- Grok 4.6: 44.5 (#66)

| Benchmark | DeepSeek-V3 | Grok 4.6 |
|---|---|---|
| LMArena Longer Query | 1352 | 1454 |
| Fiction.LiveBench | 50% | — |

## Writing & Preference

- DeepSeek-V3: 57.4 (#130)
- Grok 4.6: 62.3 (#80)

| Benchmark | DeepSeek-V3 | Grok 4.6 |
|---|---|---|
| LMArena Text | 1375 | 1428 |
| LMArena Creative Writing | 1364 | 1428 |
| LMArena Multi-Turn | 1389 | 1425 |
| Short-Story Creative Writing | 77% | — |
| EQ-Bench Creative Writing | 1472 | — |
| WildBench | 83% | — |
| LiveBench Language | 49.1% | — |

## FAQ

### Is DeepSeek-V3 better than Grok 4.6?

Grok 4.6 is the stronger model overall, scoring 56.9 to 39.5 on the Noometry Index. DeepSeek-V3 costs 7.4× less per token, which makes it the better buy when Grok 4.6's lead doesn't matter for your workload.

### Which is cheaper, DeepSeek-V3 or Grok 4.6?

DeepSeek-V3 is cheaper. It lists at $0.24 per million input tokens and $0.90 per million output tokens; Grok 4.6 lists at $2 and $6.

### Is DeepSeek-V3 or Grok 4.6 better for coding?

Grok 4.6 scores higher on coding benchmarks: 58.5 versus 42.3 in the Noometry coding category.

### Which has the bigger context window?

Grok 4.6 does, with 500K tokens against 164K.

### How many benchmarks do DeepSeek-V3 and Grok 4.6 share?

26 benchmarks have published results for both models. DeepSeek-V3 has 60 scored results on Noometry and Grok 4.6 has 49.